Anthropic Files for IPO in Confidential Submission

AI company Anthropic has confidentially submitted a draft S-1 registration statement to the SEC, and may proceed with an IPO after review.

@AnthropicAI

Anthropic has taken a decisive step toward becoming a publicly traded company. The AI lab confidentially filed a draft S-1 with the U.S. Securities and Exchange Commission, a move that could make it one of the most closely watched technology IPOs in recent years. Pending completion of the SEC review process, the company may proceed with an initial public offering, marking a major milestone for the AI industry and signaling growing investor appetite for frontier AI companies. The confidential nature of the filing means financial details remain undisclosed, but the submission itself underscores Anthropic's rapid maturation from a research-focused lab into a major commercial enterprise. With flagship Claude models competing at the frontier of language intelligence, an IPO could provide the capital needed for the next phase of AI infrastructure scaling and model development.

MiniMax Launches M3: First Open Model Combining Three Frontier Capabilities

MiniMax M3 achieves 59.0% on SWE-Bench Pro, supports 1M token context, and uses sparse attention to keep inference costs low.

@MiniMax_AI

MiniMax stunned the open-source community with the release of M3, a model that combines frontier-level coding, one-million-token context windows, native multimodality, and a novel sparse attention architecture in a single open-weights package. On SWE-Bench Pro, M3 scored 59.0%, placing it among the best coding models available. Terminal Bench 2.1 results reached 66.0%, SWE-fficiency hit 34.8%, and on KernelBench Hard it scored 28.8%. MCP Atlas performance was 74.2%. What sets M3 apart is MiniMax Sparse Attention, which scales context to one million tokens efficiently without the quadratic cost of traditional attention mechanisms. The model is available day-one on Vercel AI Gateway, Ollama Cloud, Cloudflare AI Gateway, and multiple other platforms, with launch partners offering promotional pricing. Early benchmarks place M3 behind only Claude Opus 4.8 and GPT-5 on Next.js agent evaluations, but at roughly one-tenth the cost per token.

OpenAI Frontier Models Now on AWS Bedrock for Enterprise

OpenAI's frontier models and Codex are now available for enterprises via Amazon Bedrock, supporting secure and compliant workflows.

@OpenAI

OpenAI announced that its frontier models and Codex are now generally available on AWS, giving enterprises a new way to build on Amazon Bedrock with OpenAI through the security, compliance, and governance workflows they already use. This marks the beginning of a broader expansion of OpenAI's enterprise footprint. Developers can now build AI applications and software engineering workflows with OpenAI models using the AWS environments and controls their teams already trust, bridging the gap between cutting-edge AI capabilities and enterprise-grade infrastructure.

Runway Establishes London European HQ, Invests $100 Million

Runway announced a London European headquarters and new world model research center, planning $100M investment over 18 months and expansion through 2028.

@runwayml

Runway announced London as its new European headquarters and newest research hub focused on general world models. Over the next 18 months, the company plans to invest $100 million into the UK AI ecosystem, with that figure more than doubling through 2028 as European operations scale. The move signals growing European investment in physical AI and world model research, with London positioned as a key hub for frontier AI development outside North America.